Stability Control Improvement and State Detection for Autonomous Vehicles

The proposed research will look at a combination of low-cost radar, stereo camera, and lidar data to generate a 3D model of the terrain ahead of a vehicle. The resulting terrain model will be automatically segmented, simplified and compressed to produce a reasonable description, which can interact with a vehicle model in real time to generate realistic forces on tires. The resulting estimated vehicle path will be examined for rollover or other hazard occurrence for driver alert. Further, the path will be compared with an ideal path, and a cost function will be applied to determine a measure of its"goodness."This measure will be used in a Model Predictive Control Optimizer to automatically modify autonomous steering, brake, and throttle inputs in a simulated UGV. Simulations will be run in CarSim and TruckSim to verify the method. The simulation will interact with a controller that runs in Matlab.